emodnet thematic lot n° 4/si2.749773
Post on 14-May-2022
3 Views
Preview:
TRANSCRIPT
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry Regional climatologies produced with Data-
Interpolating Variational Analysis (DIVA). Release 2018
L. Buga, G. Sarbu (NIMRD), K. Eilola, K. Wesslander, L. Fryberg, W. Magnus (SMHI), J. Gatti, D. Leroy
(IFREMER), S. Iona, M. Tsompanou, P. Karagevrekis (HCMR), Jonas Koefoed Rømer, Larsen (AU-DCE), A.K.
Østrem (IMR), M. Lipizer, A. Giorgetti (OGS)
Date: April, 2019
Citation: L. Buga, G. Sarbu, K. Eilola, K. Wesslander, L. Fryberg, W. Magnus, J. Gatti, D. Leroy, S. Iona, M.
Tsompanou, P. Karagevrekis, Jonas Koefoed Rømer, Larsen, J. Koefoed Rømer, A.K. Østrem, M. Lipizer, A.
Giorgetti. 2019 EMODnet Chemistry Regional climatologies produced with Data-Interpolating Variational Analysis
(DIVA). Release 2018. Doi 10.6092/A8CFB472-10DB-4225-9737-5A60DA9AF523
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
2
Content:
1. Introduction ........................................................................................................................................ 1
2. Mediterranean Sea ............................................................................................................................. 3
3. Black Sea ............................................................................................................................................. 6
4. Arctic region ...................................................................................................................................... 12
5. Baltic Sea ........................................................................................................................................... 14
6. North Sea .......................................................................................................................................... 19
7. North East Atlantic Ocean ................................................................................................................. 22
8. References......................................................................................................................................... 27
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
1
1. Introduction
The previous step resulted in harmonised, aggregated and validated data collections for specific
parameters, which were considered as first line data products. The next step was aimed at using these
validated data collections as input for preparing spatially interpolated concentration basin maps, with
DIVA (Data-Interpolating Variational Analysis) software.
EMODnet Chemistry Regional climatologies have been produced for the parameters listed in Tab. 1.1, at
seasonal scale and for periods of 6 years, using a moving 6-year window. The season definitions used
are defined specifically for each basin. 6-year periods span from 1980-1985 until 2013-2018. Depth range
(IODE standard depths): -300.0, -250.0, -200.0, -150.0, -125.0, -100.0, -75.0, -50.0, -40.0, -30.0, -20.0, -
10.0, -5.0, -0.0. Data Sources: observational data from SeaDataNet/EMODnet Chemistry Data Network.
EMODnet Chemistry Regional climatologies have been produced using DIVA (Data-Interpolating
Variational Analysis) software which allows to spatially interpolate sparse and irregularly distributed
observations on a regular grid in an optimal way. DIVA allows also to compute error maps for the gridded
fields which reflect the accuracy of the observations and their distribution and allow to assess the
reliability of the gridded fields.
To produce horizontal maps, profiles were interpolated at standard depths using weighted parabolic
interpolation algorithm (Reiniger and Ross, 1968). GEBCO 1min topography is used for the contouring
preparation. EMODnet Chemistry Regional climatologies are provided as masked using relative error
threshold 0.3 and 0.5. DIVA settings: signal-to-noise ratio and correlation length were estimated using
data mean distance as a minimum (for L) and vertically filtered.
Table 1.1: List of variables used to produce EMODnet Chemistry Regional climatologies and related
measurement units.
Variable Units
Ammonium umol/l
Dissolved inorganic nitrogen (DIN) umol/l
Phosphate umol/l
Silicate umol/l
Chlorophyll - a mg/m3
Dissolved oxygen umol/l
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
2
For each of the following sea basins: Arctic region, Atlantic Sea, Baltic Sea, Black Sea, Mediterranean
Sea and North Sea, seasonal climatological maps built as 6-years running averages are available (Fig. 1).
Figure 1.1: EMODnet Chemistry Regional climatologies.
For each Regional climatology, following WMS layers are available (Fig. 1.2):
Water body gridded map of the parameters listed in Tab. 1, masked using relative error threshold 0.5
Interpolated observations: station positions used to build the Water body gridded map
Additional fields:
Water body gridded map Error standard deviation of the gridded map Water body gridded map masked using relative error threshold 0.3 Relative error of gridded map Logarithm10 of number of data in bins Logarithm10 of number of outliers data in bins Correlation length field Deepest value of water body gridded map Deepest value of water body gridded map masked using relative error threshold 0.3 Deepest value of water body gridded map masked using relative error threshold 0.5
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
3
Figure 1.2: WMS layers available for Regional climatologies.
Basin specific details on DIVA map production are listed in the sections below.
2. Mediterranean Sea
2.1 General Settings
The Mediterranean climatology is defined between 7.000◦W-36.375°E of longitude and 30.000-45.875°N
of latitude. The horizontal resolution is 0.125°. Seasonal DIVA maps were produced on 6-years running
averages (starting year depending on data availability). Seasons definition for Mediterranean Sea are:
Winter (January – March), Spring (April – June), Summer (July – September) and Autumn (October –
December). DIVA maps were produced for the IODE standards depth levels, depending on data
availability for every parameter.
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
4
DIVA 4.7.2 settings and driver file input specifications., considered for both analysis and reference field
computation, are listed in Table 2.1. Default value for the correlation length is fixed to 2° in the param
file, however the correlation lengths were optimized as follows: for each parameter and for the whole
period of analysis, four seasonal vertically filtered profiles were computed which were averaged to a
mean vertical profile. This mean vertical profile was used all at the 6-years analysis. The signal to noise
ratio have been set to 1 for all computations.
Analysis has been made using the mean field as background (ireg=1 in param file) and a
logarith/exponential transformation (analysis = 11 in driver file) in order to avoid negative values in the
concentration fields. No detrending of data and no advection constraints were used. Only observations
with a quality flag equals to 1, 2, and 6 have been retained for the maps’ generation with DIVA
Table 2.1 – DIVA configuration used for the Mediterranean products.
Param file Driver file
Lc: correlation length 2.0 (default value) Data extraction 1
icoordchange 1 Boundary lines and coastline generation
1
ispec 113 Cleaning data on mesh 1
ireg 1 Minimal number of data in a layer
-1
xori -7.0 Parameters extimation and vertical filtering
-30
yori 30.0 Minimal L 0.1
dx 0.125 Maximal L 10
dy 0.125 Mininal SN 0.1
nx 348 Maximal SN 5.0
ny 128 Analysis and reference field
11
valex -9999.9 Lower level generation 1
snr 1.0 Upper level generation 33
varbak 1.0 4D netcdf file generation
11
Gnuplots 0
Data detrending 0
2.2 Results
For data description and numbers, please refer to the Aggregated datasets documentation and/or to
Sextant Descriptions.
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
5
All maps produced for Mediterranean Sea are detailed in Table 3.2 2.2. The number of depth levels as
well as the temporal extent vary between the parameters. The choice of the deepest layers was based
on a minimum number of data in a layer (from 5 to 10 values) in combination with a minimum number
of 6-year periods having that minimum number of data in the layer.
Table 2.2 - Availability of DIVA maps for the Mediterranean Sea for each parameter and as a function
of season and depth.
Parameter Availability
Depths Winter season
Spring season
Summer season
Autumn season
Water body dissolved oxygen
27 levels from surface to 2000 m deep
from 1971-2017
Water body phosphate
25 levels from surface to 1500 m deep
from 1968-2016
Water body silicate
25 levels from surface to 1500 m deep
from 1965-2016
Water body dissolved inorganic nitrogen (DIN)
21 levels from surface to 1100 m deep
from 1990-2016
Water body chlorophyll-a
20 levels from surface to 1000 m
from 1990-2017
2.3 Validation
The 6-years DIVA maps were evaluated only by direct comparisons of the maps with the observations
used for their preparation because of lack of a similar product for the Mediterranean of the same
temporal resolution (6-years). Some additional visual checks were performed between the DIVA maps
and the World Ocean Atlas 2013. Although the two products have different spatial and temporal
resolution, the comaprison was done for evaluation of the general characteristics and features a=in the
region. More specific there were compare the 6-years DIVA maps for oxygen, phosphate and silicate
with the seasonal World Ocean Atlas 2013 climatology of 1ox1o resolution
(https://www.nodc.noaa.gov/OC5/SELECT/woaselect/woaselect.html). As the Mediterranean DIVA
productis of higher spatila resolution , it reveals in more details the general features of the seasonal
distribution of the parameters.
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
6
3. Black Sea
3.1 General Settings
The Black Sea, the world’s largest anoxic basin, is a deep ocean basin (~2000 m) connected to the
Mediterranean Sea by the narrow Bosporus Strait, Marmara Sea, and Dardanelles. The only source of
salty water is from the Mediterranean through the Bosporus Strait via the Sea of Marmara. As a result,
there is a strong salinity-driven density gradient and a permanent halocline which limits vertical supply
from the deep. This has resulted in permanent anoxia in waters below 125 – 200 m over most of the
basin (Oguz et al., 2008).
DIVA analysis domain definition
The MSFD Black Sea domain with boundaries 40N – 48N, 26.5E – 42E includes Black Sea and the adjacent
Sea of Azov at the NE (connected to the Black sea via Kerch strait) and Marmara Sea at SW (connected
to the Black sea via Bosporus strait) (Figure 3.1).
Figure 3.1. The Black Sea MSFD domain
Time-space resolution
Seasonal DIVA maps were produced on 6-years running averages (starting year depending on data
availability). Seasons definition for Black Sea region are: Winter (December – February), Spring (March
– May), Summer (June – August) and Autumn (September – November). Considering the data coverage
presented in Black Sea, DIVA maps were produced for the IODE standards depth levels, depending on
data availability for every parameter. The horizontal resolution of the produced DIVA maps grids is 0.05
degrees.
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
7
DIVA (version 4.7.1) settings and driver file input specifications., considered for both analysis and
reference field computation, are listed in Table 3.1.
Due to the high variability of data coverage in time and space, correlation length and signal-to-noise
ratio parameters are estimated (per each season, each 6-years period and each vertical level) using data
mean distance as a minimum (for L), and both parameters are vertically filtered (parameters estimation
and vertical filtering = -30 in driver file). Analysis has been made using the seasonal (per each 6-years
period) mean field as background (ireg=1 in param file) and a logarith/exponential transformation
(analysis = 11 in driver file) in order to avoid negative values in the concentration fields.
No detrending of data, no reference field and no advection constraints were used. Accepted quality flags:
1 ,2, 5, 6, 8 (SeaDataNet scheme http://seadatanet.maris2.nl/v_bodc_vocab_v2/search.asp?lib=L20).
Table 3. 1 – DIVA configuration used for the Black sea products
Param file Driver file Lc: correlation length 1.5 Data extraction 1 icoordchange 2 Boundary lines and coastline generation 1 ispec 113 Cleaning data on mesh 4 ireg 1 Minimal number of data in a layer 10 xori 2.5 Parameters estimation and vertical filtering -30 yori 40 Minimal L 0.2 dx 0.05 Maximal L 10 dy 0.05 Mininal SN 0.1 nx 360 Maximal SN 5.0 ny 160 Analysis and reference field 11 valex -99 Lower level number 1 snr 0.5 Upper level number Depending on
parameter varbak 1.0 4D netcdf file generation 11
Gnuplots 0
Data detrending 0
3.2 Results
The Black Sea seasonal on 6-years running averages DIVA maps were produced from the aggregated
Eutrophication Black Sea Dataset containing 54894 CDIs (of which 8278 are restricted) as vertical
profiles. The dataset (Quality control applied, data availability, etc) is described in EMODnet Chemistry
Eutrophication and Ocean Acidification aggregated datasets v2018. The unrestricted dataset is available
for downloading at: Black Sea - Eutrophication and Ocean Acidification aggregated datasets 1935/2016
v2018
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
8
The seasonal produced DIVA maps, the 6-years periods, as well as the number of vertical levels are
summarized in Table 3.2. The starting year for analyses is depending on data availability, for dissolved
oxygen, phosphate and silicate the first 6-year period being 1970-1975 while for chlorophyll-a and DIN
the analyses started with 1990-1995. The number of depth level also depend on data availabilty and for
the Black Sea there too few data below 200-250 m deph for the DIVA analyses. Some examples of DIVA
maps are illustrated in Figure 3.2.
Table 3.2 – Availability of DIVA maps for the Black Sea Region
Parameter Availability
IODE Depth levels
Winter Spring Summer Autumn
Water body dissolved oxygen 13 levels,
0, 5, 10, …, 250 m 1970-1975, 1971-1976, ....., 2011-2016
Water body dissolved inorganic nitrogen (DIN)
9 levels, 0, 5, 10, … 100 m
1990-1995, 1991-1996, ..., 2011-2016
Water body phosphate 12 levels,
0, 5, 10, …, 200 m 1970-1975, 1971-1976, ....., 2011-2016
Water body silicate 12 levels,
0, 5, 10, …, 200 m 1970-1975, 1971-1976, ....., 2011-2016
Water body chlorophyll-a 7 levels,
0, 5, 10, …, 50 m 1990-1995, 1991-1996, ....., 2011-2016
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
9
Figure 3.2. Examples of
DIVA masked using relative error
threshold 0.5 maps:
a) Water body dissolved oxygen,
summer, 1989-994, 75 m depth
b) Water body chlorophyll-a, spring,
2003-2008, 0 m depth
c) Water body phosphate, winter,
1997-2002, 0 m depth
Black Sea DIVA products validation
Due to the specificity of eutrophication DIVA products, the EMODnet seasonal 6-year running averages
climatologies can not be directly compared with other climatologies like World Ocean Atlas (WOA). Even
more, the spatial resolution of WOA products for nutrients (oxygen, phosphate an silicate) is 1°, a
resolution too low to reveal the characteristics of nutrients distribution over the entire Black Sea basin
and specially around the coastal areas.
a)
b)
c)
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
10
Black Sea high resolution eutrophication DIVA products
The Black Sea is supplied by a number of major rivers, such as the Danube, Dnieper, Southern Bug,
Dniester, Don, and Rioni (Figure 3.3). The Danube River is the major source of fresh water in the Black
Sea Basin. Its discharge is estimated at ~200 km3/year (6000 m3/s). The Danube waters are an important
source of nutrients, providing approximately 200–500 kt/year of dissolved inorganic, nitrogen, ~15–30
kt/year of phosphates, ~200–400 kt/year of silicates (Kubryakov et al., 2018)
Figure 3.3 Drainage and schematic water circulation of the Black Sea
The region in front of Danube Delta months (black circle and excerpt in Figure 3.3) was chosen for
analyses with DIVAnd to produce high resolution maps for water body phosphate, DIN and silicate.
High resolution maps were produced for the domain 28-30°E, 43.7-46.25°N using Gebco 30sec
bathymetry, for the depths from 2.5 to 2.5 m from 0 up to 60 m. A horizontal correlation length of
60 km and a vertical correlation length of 10 m were applied. The horizontal resolution of the produced
DIVAnd maps grids is 0.005 degrees. The 6-year running averages maps were produced for all seasons,
starting with 1990.
Example of DIVAnd maps are shown in Figure 3.4.
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
11
Figure 3.4 High resolution DIVAnd maps, 2006-2011, summer, 12.5 m depth. The observational data are shown in the upper panel.
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
12
4. Arctic region
4.1 General Settings
The Arctic region covers the Norwegian Sea, Barents Sea, Arctic Ocean, Greenland Sea and Icelandic
Waters, defined by ICES Ecoregions since there is no MSFD definition for this region. The domain extends
from 56-83°N, and from 45°W - 70°E, with horizontal resolution 0.25° x 0.125°. IODE standard depths
from surface to 3000 m are used for 3 of the parameters, and down to 300 m for chlorophyll-a. Time
period used is 1960-2017, except for chl-a that starts in 1980. Seasons definitions for the region are
winter (December-February), spring (March-May), summer (June-August) and autumn (September-
November).
Climatologies were produced using DIVA version 4.7.2, with the settings listed in Table 4.1. Correlation
length and signal-to-noise ratio parameters are estimated using data mean distance as a minimum (for
L), and both parameters are vertically filtered. Analysis is performed with log(data) - exp(analysis)
transformation (option 11) to avoid negative values in the concentration fields. A reference field with all
seasonal data for the whole period (1960-2017; 1980-2017 for chl-a) was used as background field.
Table 4.2 – DIVA configuration used for the Arctic region products.
Param.par Driver
Lc: correlation length 1.0 Data extraction 1
icoordchange 2 Boundary lines and coastlines generation 1
ispec 113 Cleaning data on mesh 1
ireg 0 Minimal number of data in a layer 0
xori -45 Parameters estimation and vertical filtering -30
yori 56 Minimal L 1.0
dx 0.25 Maximal L 5.0
dy 0.125 Minimal SN 0.1
nx 461 Maximal SN 10
ny 217 Analysis and reference field 11
valex -9999.0 Lowerlevel number 1
snr 0.5 Upperlevel number 30
varbak 1.0 4D netCDF files generation 11
Gnuplot plots 1
Data detrending 0
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
13
4.2 Results
The Arctic eutrophication dataset contains 43 844 CDIs. Of this, only 844 CDIs are restricted data, the rest are unrestricted. The dataset is described in the report EMODnet Chemistry Eutrophication and Ocean Acidification aggregated datasets v2018. The unrestricted dataset is available for downloading at http://www.emodnet-chemistry.eu/products/catalogue#/metadata/35828f42-27f0-4b05-ac3b-f3d77632e180. In addition to the Arctic dataset, an overlapping dataset from the North Sea was included in the DIVA analysis to minimize the potential interpolation differences near the border between the two regions. After analysis the overlapping region was cut off, using the ICES Ecoregions boundaries. Table 4.2 lists the DIVA maps produced for each season, the number of depth layers, and the first and last 6-year periods. For each parameter, the same number of layers and years have been used for all seasons. Since the Arctic dataset contains only 37 profiles of ammonium, it was not possible to produce DIVA maps for ammonium or DIN. Table 4.2 – Availability of DIVA maps per season for the Arctic region, including depth layers and years
Parameter Availability
Depth layers
Winter (Dec-Feb)
Spring (Mar-May)
Summer (Jun-Aug)
Autumn (Sep-Nov)
Water body dissolved oxygen
30 layers, 0-3000 m
1960-1965, ... , 2012-2017
Water body phosphate 30 layers, 0-3000 m
1960-1965, ... , 2012-2017
Water body silicate 30 layers, 0-3000 m
1960-1965, ... , 2012-2017
Water body chlorophyll-a
14 layers, 0-300 m
1980-1985, ... , 2012-2017
DIVA sometimes gives very high values in fjords and bays, even when no data is available there. In the
White Sea the values were sometimes as much as 100 times higher than the data values, and the results
in this area were therefore removed from the final products. The results in the fjords were kept because
there is not enough data to produce maps in the winter season without it. The distribution of data varies
a lot between seasons, years and parameters, making it difficult to find settings that work in all cases.
After publication of the maps, it was discovered that some of them contained NaN's (in some years or
depths), particularly for chlorophyll-a in spring. Tests show that this probably has to do with the use of
reference fields, but the reason is not clear.
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
14
Some tests have been done with higher resolution near river mouths. In this region there are no large
rivers, and the river mouths are usually in a fjord, not at the coast. In the fjords there is data only in
November/December, and no data after 2008. The tests were performed with DIVA 4.7.2, and the results
in the fjords are not good. The new version of DIVA has not been tested with higher resolution yet.
5. Baltic Sea
5.1 General Settings
The Baltic Sea is bordered by nine countries with about 85 million inhabitants (Fig. BS1). The Baltic Sea
is a semi-enclosed, brackish and shallow sea with limited water exchange with the open ocean. The
freshwater supply from rivers causes large salinity gradients between the northern sub-basins and the
entrance area and between the well-mixed surface layer and the saltier bottom layer, respectively.
Enhanced eutrophication since World War II has led to increased frequency and intensity of
cyanobacterial blooms and severe oxygen deficiency with expanding anoxic and dead bottom areas in
the Baltic Sea.
5.2 Baltic Sea DIVA eutrophication products
The entire set of Baltic Sea DIVA maps can be found in the OceanBrowser viewing service:
http://ec.oceanbrowser.net/emodnet/.
The data products are based on climatological (6-years running averages) maps of selected parameters
(BS Table 1). The interpolated DIVA data set with corresponding method and metadata descriptions is
available for download from the EMODnet Chemistry Sextant Catalogue Service http://www.emodnet-
chemistry.eu/products/catalogue#/search?from=1&to=20.
Table 5.1. The Baltic Sea DIVA products related to eutrophication. The 6-year running averages of the
parameters are shown as maps at discrete depth layers and there is also a map of the deepest values
available.
Region Season Water body parameter
Baltic Sea Autumn (Sep-Nov)
Spring (Mar-May)
Summer (Jun-Aug)
Winter (Dec-Feb)
ammonium, chlorophyll-a, DIN, oxygen, phosphate, silicate
ammonium, chlorophyll-a, DIN, oxygen, phosphate, silicate
ammonium, chlorophyll-a, DIN, oxygen, phosphate, silicate
ammonium, chlorophyll-a, DIN, oxygen, phosphate, silicate
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
15
The production of DIVA maps for the Baltic Sea is generally described at http://www.emodnet-
chemistry.eu/products/generation and has followed the basic guide lines for DIVA maps production of
EMODnet Chemistry http://www.emodnet-
chemistry.eu/phase2/default/products/documents/guidelines_DIVA_maps_production.pdf.
The Baltic Sea aggregated Eutrophication data set used for the production of DIVA maps included in total
187 597 CDI:s (vertical profiles). The aggregated data set with corresponding method and metadata
descriptions is available for download from the EMODnet Chemistry Sextant Catalogue Service
http://www.emodnet-chemistry.eu/products/catalogue#/search?from=1&to=20. In addition to these
open data a dataset with restricted data containing in total 5183 profiles were included in the DIVA
analysis. To minimize the potential interpolation differences near the border between the North Sea and
the Baltic Sea an overlap dataset from the North Sea was included in the DIVA analysis. After the analysis
the overlap region was cut off leaving only the Baltic Sea area in the final products. The DIVA products
correspond to the 6-years running average at the standard depths (-300.0, -275.0, -250.0, -225.0, -200.0,
-175.0, -150.0, -125.0, -100.0, -90.0, -80.0, -70.0, -60.0, -50.0, -40.0, -30.0, -20.0, -15.0, -10.0, -5.0, 0.0)
of the regional convention HELCOM (Baltic Marine Environment Protection Commission - Helsinki
Commission). The horizontal correlation length is 0.7 degrees (about 80 km). Correlation length is
generally chosen to be independent of longitude and latitude but a spatially dependent field could be
provided in future work. The horizontal resolution is 0.1°. As an example a screen shot of winter water
body phosphate is shown in Fig.5.1.
Figure 5.1. EMODnet Chemistry Baltic Sea DIVA product (screen shot) winter (Dec-Feb) period 1997-
2002. 6-years running average winter water body phosphate mol l-1 (0m depth). The information
square seen in the center points out the position of monitoring station BY15 in the Gotland Deep and
the value of the DIVA product at that position. The red square indicates the Gulf of Riga.
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
16
5.3 Evaluation of Baltic Sea DIVA eutrophication products
The possibility to evaluate the Baltic Sea DIVA products against the World Ocean Atlas was investigated.
The World Ocean Database (WOD) was updated in 2018 and the related World Ocean Atlas (WOA) which
includes climatological fields and statistic data was also updated at the time. The climatological fields
with nutrients which are presented in the WOA are however only on a global scale with the resolution
1°. The resolution is unfortunately too low to be able to compare the Baltic Sea region with the EMODnet
DIVA products. The regional climatological fields that are available in the WOA have a resolution of 1/4°
or 1/10° but show only fields of salinity and temperature. Hence, an evaluation relative to the WOA was
not relevant for the Baltic Sea DIVA products.
In comparison with the DIVA product extracted at the Gotland Deep BY15 the underlying data set from
the Sextant Catalogue service was downloaded and analyzed in the Ocean Data View (ODV) software.
The data set includes only unrestricted data while the DIVA products are produced from a slightly larger
data set including also restricted data reported by the different institutes. For the parameter analysis
time series of winter (December, January, and February) dissolved phosphate (PO4) and silicate (Si) as
well as dissolved inorganic nitrogen (DIN, the sum of Nitrates, Nitrites and Ammonium) were selected.
In addition the summer (June, July and August) chlorophyll-a (Chl-a), and autumn (September, October,
November) deep water oxygen were evaluated (Fig.BS2).
The results show that the DIVA product and the observations at BY15 in the central Baltic Sea (Fig.BS1)
are well in accordance with each other. The DIVA product has the advantage that it can produce
estimates of data in periods and areas where data is missing. DIVA works as a filter in space and removes
essentially small-scale variations. This is necessary because the small-scale variations are not
appropriately sampled over the whole basin. The output of DIVA is effectively a space and time average
and in Fig.BS2 we compare it only to time averaged value.
Suggestions of future work in areas with anoxic conditions could include hydrogen sulfide as negative
oxygen. Also the uncertainty (e.g. 95% confidence bounds) of the six year running mean value of the
DIVA product could be calculated from the six years data used in the calculation. Another measure of
uncertainty beside the DIVA error could be to calculate and report the statistical distance (e.g. the bias
in mean values and standard deviation) from the corresponding values calculated from observations at
stations with many data.
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
17
Figure 5.2. Average BY15 surface layer (0m) winter water body phosphate (upper left), DIN (upper right)
and silicate (middle left) and summer chlorophyll-a (middle right) from the ODV data set (open blue
circles). The autumn oxygen concentrations at 100m depth (lower left) and 200m depth (lower right) are
also shown. The six year running average starting in year three of the period is shown by the black solid
line. The dotted black lines indicate the range of ± 1 standard deviation for the corresponding six year
period of data. The red line shows the results from the DIVA interpolation product obtained from the
EMODnet Chemistry OceanBrowser.
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
18
5.4 Baltic Sea high resolution eutrophication DIVA products
A preliminary evaluation of the availability of nutrient data from the river mouths close to the major
freshwater tributaries in the Baltic Sea generally indicated relatively poor nutrient data cover. Hence, an
increased resolution compared to the DIVA products already produced would not add new information.
The study indicated, however, that a high resolution mapping of the area in the vicinity of river Daugava
in the Gulf of Riga could be possible. The high resolution maps for the area Lon E22.3-25.0 and Lat N56.8-
N58.5 were produced with DIVAnd using bathymetry Gebco 30sec, a resolution of 0.01 and a horizontal
correlation length of 25 km for depth levels 0, -2, -5, -10, -15, -20, -30, -40, -50, meter. The six years
running mean maps were produced for water body phosphate, DIN and silicate for the years 2008 to
2013. An example from the period 2010-2015 is shown in Fig.5.3.
Figure 5.3. Six year mean (2010-2015) winter surface water body phosphate (left), DIN (middle) and
silicate (right) in the Gulf of Riga. The underlying profiles data are shown in the upper panel with the
corresponding high resolution map in the lower panels. The major river Daugava is situated in the
southern Gulf of Riga as indicated by the elevated nutrient concentrations in the maps. Concentrations
are shown in the colour bar with units in mol l-1.
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
19
6. North Sea
The EMODNet North Sea domain stretches from the Kattegat between Denmark and Sweden and out
into the open Greater North Sea and Celtic Sea encompassing the coastlines of south-mid Norway, The
British Isles, north-west Germany, The Netherlands, Belgium and north-west France (Fig. 6.1). At the
border between the North Sea and the Baltic Sea, between Denmark and Sweden, the brackish waters
of the Baltic meets the saline waters of the Kattegat and the North Sea with a strong gradient.
Figure 6.1 The North Sea EMODnet region.
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
20
North Sea DIVA eutrophication products
The entire set of North Sea DIVA maps can be found in the OceanBrowser viewing service:
http://ec.oceanbrowser.net/emodnet/.
The data products are based on climatological (6-years running averages) maps of selected parameters
(Table 6.1). The interpolated DIVA data set with corresponding method and metadata descriptions is
available for download from the EMODnet Chemistry Sextant Catalogue Service http://www.emodnet-
chemistry.eu/products/catalogue#/search?from=1&to=20.
Table 6.1. The North Sea DIVA products related to eutrophication. The 6-year running averages of the
parameters are shown as maps at discrete depth layers and there is also a map of the deepest values
available.
Region Season Water body parameter
North Sea Autumn (Sep-Nov)
Spring (Mar-May)
Summer (Jun-Aug)
Winter (Dec-Feb)
chlorophyll-a, DIN, oxygen, phosphate, silicate
chlorophyll-a, DIN, oxygen, phosphate, silicate
chlorophyll-a, DIN, oxygen, phosphate, silicate
chlorophyll-a, DIN, oxygen, phosphate, silicate
The production of DIVA maps for the North Sea is generally described at http://www.emodnet-
chemistry.eu/products/generation and has followed the basic guide lines for DIVA maps production of
EMODnet Chemistry http://www.emodnet-
chemistry.eu/phase2/default/products/documents/guidelines_DIVA_maps_production.pdf.
The North Sea aggregated Eutrophication data set used for the production of DIVA maps included in
total approximately 4.7 million CDI:s (vertical profiles). Of these, approximately 500.000 nutrients
datasets and 3 million “others” (i.e. Chlorophyll and oxygen profiles and spot samples) datasets had
relevant information for the DIVA maps. The aggregated data set with corresponding method and
metadata descriptions is available for download from the EMODnet Chemistry Sextant Catalogue Service
http://www.emodnet-chemistry.eu/products/catalogue#/search?from=1&to=20.
To minimize the potential interpolation differences near the border between the North Sea and the
Baltic Sea an overlap dataset from the Baltic Sea was included in the DIVA analysis. After the analysis the
overlap region was cut off leaving only the North Sea area in the final products. The DIVA products
correspond to the 6-years running average at the depths (-500.0, -100.0, -75.0, -50.0, -25.0, -20.0, -10.0,
-5.0, -1.0, 0.0). The horizontal correlation length is 0.7 degrees (about 80 km). Correlation length is
generally chosen, based on experiments and experience, to be independent of longitude and latitude
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
21
but a spatially dependent field could be provided in future work. Importantly, the correlation length as
well as signal to noise ratio has been aligned with the settings of the Baltic Sea, to make the products
standardized and comparable. The horizontal resolution is 0.1°.
Example and evaluation of North Sea DIVA eutrophication products
In Fig. 6.2 an example of winter phosphate is shown and compared with the adjoining sea regions.
Figure 6.2 The interpolated product of winter phosphate, masked with error level 0.5, year 1997-2002 is
shown. The DIVA products for the North Sea, Baltic Sea, Arctic Sea and Atlantic Ocean have all been
added to the view, with same symbology scale. Between the North Sea and Baltic Sea, the interpolated
results align seamlessly, and also towards The Arctic Sea the values align well. Against the Atlantic it is
harder to tell in this case, since low sample availability has caused the error level in this part of the North
Sea domain to be too high and thus masked out of the final result.
In general, most of the DIVA products aligned well with the neighbouring sea regions. However, in some
cases the switch to six-year intervals from ten years seems to have triggered a problem in the Diva
software, that caused too high values in the interpolated results. We would have expected these to be
masked out by the error level threshold, but they were not. This problem was first discovered, pointed
out to us and explored by the Baltic Sea regional leaders. The issue is noted and a solution will be pursued
in a continuation of the EMODNet project, if revisions of the products are made.
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
22
7. North East Atlantic Ocean
7.1 General Settings
The North East Atlantic Ocean domain extends between 42°W-0°W and between 24°N-49°N. Horizontal
resolution is 0.1°. Regarding seasons, winter spans from January to March, spring from April to June,
summer from July to September and Fall from October to December.
DIVA 4.7.1 settings and driver file input specifications, considered for both analysis and reference field
computation, are listed in Table 3. 1. Default value for the correlation length is fixed to 3° in the param
file, however, profiles of correlation lengths have been previously computed for each season and over
the whole periods (1960-2017) and the resulting values have been used for all 6-year climatologies. On
the opposite, signal to noise ratio have been set to 1 for all computations.
Analysis has been made using the mean field as background (ireg=1 in param file) and a
logarith/exponential transformation (analysis = 11 in driver file) in order to avoid negative values in the
concentration fields.
No detrending of data and no advection constraints were used. But, the weight of the time series was
decreased by a factor of 10 relative to the weight of the profiles to account for the redundancy in time
series observations. Only observations with a quality flag equals to 1, 2, 5, 6, 7 or 8 have been retained
for the maps’ generation with DIVA.
Table 7.1 – DIVA configuration used for the North East Atlantic Ocean products.
Param file Driver file
Lc: correlation length 3.0 (default value) Data extraction 1
icoordchange 2 Boundary lines and coastline generation
1
ispec 113 Cleaning data on mesh 1
ireg 1 Minimal number of data in a layer
0
xori -42.0 Parameters extimation and vertical filtering
-10
yori 24.0 Minimal L 0.25
dx 0.1 Maximal L 10
dy 0.1 Mininal SN 0.1
nx 420 Maximal SN 3.0
ny 250 Analysis and reference field
11
valex -99 Lower level generation 0
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
23
snr 1.0 Upper level generation 30
varbak 1.0 4D netcdf file generation
11
Gnuplots 0
Data detrending 0
7.2 Results
For data description and numbers, please refer to the Aggregated datasets documentation and/or to
Sextant Descriptions.
All maps produced for the North East Atlantic Ocean are detailed in Table 3.2. The number of depth
levels varies between one parameter to the other. For dissolved oxygen, maps have been produced from
the surface down to 3000 m deep. The 2 lowest levels have been suppressed for phosphate, silicate and
DIN leading to 28 different depths down to 2000 m. Finally, only the first 14th levels down to 300m have
been kept for chlorophyll –a.
The first 6-year period that is available depends of course on data availability and thus, on the parameter
and the season. It was only possible to produce the 52 6-year climatologies for dissolved oxygen, no
matter the season. For chlorophyll-a, all climatologies start only with the 1972-1977 period. For
phosphate, silicate and DIN, starting period depends on the season. Some maps are available from the
sixties for phosphate. Silicate only starts in the seventies and DIN in the eighties.
Table7.3 – Availability of DIVA maps for the North East Atlantic Ocean for each parameter and as a
function of season and depth.
Parameter Availability
Depths Winter season
Spring season
Summer season
Autumn season
Water body dissolved oxygen
30 levels from surface to 3000 m deep
from 1960-1965
Water body phosphate
28 levels from surface to 2000 m deep
from 1967-1972
from 1960-1965
from 1968-1973
from 1961-1966
Water body silicate
28 levels from surface to 2000 m deep
from 1972-1977
from 1984-1989
from 1972-1977
from 1971-1976
Water body dissolved inorganic nitrogen (DIN)
28 levels from surface to 2000 m deep
from 1988-1993
from 1988-1993
from 1986-1991
from 1984-1989
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
24
Water body chlorophyll-a
14 levels from surface to 300 m
from 1972-1977
7.3 Validation
Evaluating the quality of the 6-year climatologies produced by DIVA is quite challenging. Indeed,
comparing the availaible observations with DIVA results is not very relevant as the observations have
already been used by DIVA to compute the fields. Thus, observations and maps results for a given
location will obvioulsy show a good agreement.
Regarding the existing climatologies, the World Ocean Atlas includes products for temperature, salinity,
oxygen, phosphate, silicate and nitrate. However, the spatial and temporal resolution for the fields
depends on the parameter. For oxygen, phosphate and silicate, products are only available at 1° or 5°
resolution, for the 4 seasons but for all decades (i.e. from 1955 to 2017). No direct comparison is thus
possible with EMODnet 6-year climatologies. But we compared DIVA maps produced over the whole
period (1960-2017) with WOA products for oxygen, phosphate and silicate and found a general good
agreement between the two. Figure 7.11 shows an example for dissolved oxygen at 1100m depth in
winter time. DIVA field has a better resolution (0.1° compared to 1° for WOA) and thus appears smoother
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
25
but the fields’ structure is the same, showing a gradient in the concentrations from the north-west
corner to the south east corner.
Figure 7.11 – Comparison between DIVA results (on the left) and WOA 2013 V2 (on the right) for oxygen
at 1100m depth in winter. Dots on the left figure represent observations that have been used for DIVA
calculations.
8. Summary
All products are seasonal climatology based on a 6-year period.
Name of the
product
Sea Basin Year of
publication
Horizontal
resolution
Vertical levels Available variables
EMODnet
Chemistry
Regional
climatologies –
Mediterranean
Sea
Mediterranean
Sea
2018 0.125°
0.125°
• 20 levels from surface to 1000 m • 21 levels from surface to 1100 m deep • 27 levels from surface to 2000 m deep
Water body chlorophyll-a
Water body dissolved inorganic nitrogen (DIN)
Water body dissolved oxygen concentration
Water body phosphate
Water body silicate
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
26
• 25 levels from surface to 1500 m deep • 25 levels from surface to 1500 m deep
EMODnet
Chemistry
Regional
climatologies –
Black sea
Black Sea, Sea of Azov, Sea of Marmara
2018 0.05° x
0.05°
7 9 13 12 12
Water body chlorophyll-a
Water body dissolved inorganic nitrogen (DIN)
Water body dissolved oxygen concentration
Water body phosphate
Water body silicate
EMODnet
Chemistry
Regional
climatologies –
Arctic region
Arctic region 2018 0.25° x
0.125°
-3000.0, -2500.0, -2000.0, -1750.0, -1500.0, -1400.0, -1300.0, -1200.0, -1100.0, -1000.0, -900.0, -800.0, -700.0, -600.0, -500.0, -400.0, -300.0, -250.0, -200.0, -150.0, -125.0, -100.0, -75.0, -50.0, -40.0, -30.0, -20.0, -10.0, -5.0, -0.0
Water body chlorophyll-a
Water body dissolved oxygen concentration
Water body phosphate
Water body silicate
EMODnet
Chemistry
Regional
climatologies –
Baltic Sea
Baltic Sea 2018 0.1 ° -300.0, -275.0, -
250.0, -225.0, -
200.0, -175.0, -
150.0, -125.0, -
100.0, -90.0, -80.0,
-70.0, -60.0, -50.0, -
40.0, -30.0, -20.0, -
15.0, -10.0, -5.0,
0.0.
Water body chlorophyll-a
Water body dissolved inorganic nitrogen (DIN)
Water body dissolved oxygen concentration
Water body phosphate
Water body silicate
North Sea North Sea and
Celtic Sea
2018 0.1 ° -500.0, -100.0, -
75.0, -50.0, -25.0, -
20.0, -10.0, -5.0, -
1.0, 0.0
Water body chlorophyll-a
Water body dissolved inorganic nitrogen (DIN)
Water body dissolved oxygen concentration
Water body dissolved oxygen saturation
Water body phosphate
Water body silicate
EMODnet Thematic Lot n° 4/SI2.749773
EMODnet Chemistry maps produced with Data-Interpolating Variational
Analysis (DIVA)
27
EMODnet
Chemistry
Regional
climatologies –
North East
Atlantic Ocean
North East
Atlantic Ocean
2018 0.1° Down to 300m
Down to 2000m
Down to 3000m
Water body chlorophyll-a
Water body phosphate
Water body silicate
Water body dissolved inorganic nitrogen (DIN)
Water body dissolved oxygen
concentration
9. References
Oguz T, Salihoglu B, Fach B, 2008, A coupled plankton–anchovy population dynamics model assessing
nonlinear controls of anchovy and gelatinous biomass in the Black Sea, Mar Ecol Prog Ser 369:229-256.
doi.org/10.3354/meps07540
A.A. Kubryakov, S.V. Stanichny, A.G. Zatsepin, 2018, Interannual variability of Danube waters
propagation in summer period of 1992 – 2015 and its influence on the Black Sea ecosystem, Journal of
Marine Systems, Volume 179, March 2018, Pages 10-30, https://doi.org/10.1016/j.jmarsys.2017.11.001
top related